Skip to main content
A patient is comforted by her husband at St. Joseph's Hospital in Tucson, Arizona, U.S. September 13, 2017. They were enrolled in health insurance thanks to Maria Losoya, a University of Arizona Center for Rural Health Navigator. Photo taken September 13, 2017.     REUTERS/Caitlin O'Hara - RC14E813EA30
Report

Effects of weakening safeguards in the Administration’s Health Reimbursement Arrangement proposal

Editor's Note:

This analysis is part of the USC-Brookings Schaeffer Initiative for Health Policy, which is a partnership between the Center for Health Policy at Brookings and the University of Southern California Schaeffer Center for Health Policy & Economics. The Initiative aims to inform the national health care debate with rigorous, evidence-based analysis leading to practical recommendations using the collaborative strengths of USC and Brookings.

In late October, the Trump Administration released a proposed rule that would permit employers to subsidize their employees’ purchase of individual market health insurance coverage via health reimbursement arrangements (HRAs). Under current law and regulations, employers are generally barred from subsidizing individual market coverage on behalf of their employees, whether using pre-tax or post-tax dollars, although small employers are eligible for a limited exception from this rule.

As Christen Linke Young, Jason Levitis, and I discuss in a recent analysis of the proposed rule, this change would have wide ranging effects. Notably, many large employers with sicker workforces would find it attractive to subsidize individual market coverage via an HRA.1 Today, these employers’ only options for providing coverage to their workers are to purchase experienced-rated coverage on the large group market or to self-insure. Thus, access to community-rated individual market products could allow these employers to provide coverage at considerably lower cost. However, the resulting changes in coverage arrangements would likely increase individual market premiums and the federal deficit.

While the Administration’s proposal in its current form would mostly result in the type of firm-level shifting described above, employers would also have an incentive to selectively shift their sicker workers into the individual market while arranging to cover their healthier workers in some other way.2 The proposed rule includes several provisions designed to prevent this behavior: it bars employers from offering this type of HRA side-by-side with a traditional health plan; it requires employers to offer this type of HRA on the same terms to all similarly situated employees; and it bars employers from using this type of HRA to subsidize short-term, limited duration coverage. While these safeguards would likely be largely effective (although they can and should be strengthened), the proposed rule solicits comments on eliminating or weakening them.

This analysis examines how eliminating one or more of these safeguards would affect individual market premiums. I estimate that, absent these safeguards, the Administration’s HRA proposal would cause very large increases in individual market premiums, although the precise magnitude of those increases is uncertain. In a conservative scenario in which only 10 percent of employers elected to use HRAs to shift sicker workers into the individual market, individual market premiums would rise by 16 percent or more. If all employers took this approach, individual market premiums would almost double. These effects are so large because the group market is so much larger than the individual market, which means that even moderate amounts of worker-level shifting can substantially alter the individual market’s risk mix.

The remainder of this analysis proceeds as follows. I first briefly summarize the Administration’s proposal and its main effects, drawing heavily on the analysis co-authored with my Brookings colleagues referenced above. I then examine the consequences of relaxing or eliminating the safeguards against worker-level shifting that were included in the proposed rule and present quantitative estimates of the effects of a version of the Administration’s HRA proposal that lacked one or more of these safeguards.

Overview of the Proposed Rule

An HRA is a mechanism by which an employer can pay for certain health care expenses incurred by its employees on a pre-tax basis. Currently, employers are generally prohibited from subsidizing an employee’s purchase of individual market coverage, whether through an HRA or other mechanisms. (In 2016, Congress created a limited exception to this rule for employers with fewer than 50 full-time-equivalent employees.) The proposed rule would permit HRAs that can be used to pay premiums for individual market coverage under certain conditions discussed below.3

Under the proposed rule, this new option would likely appeal primarily to large employers with sicker workforces. When offering a traditional health plan, large employers must either self-insure or purchase coverage in the large group market, where premiums are typically experience-rated; either way, the cost of providing coverage reflects the health status of their employees. By contrast, individual market premiums are community-rated, meaning they do not vary by health status. Offering an HRA that can be used to purchase individual market coverage could therefore allow employers with sicker workforces to offer similar coverage at much lower cost. (The small group market, like the individual market, is community-rated, so the proposed rule would create no such benefits for small employers.)

These incentives would likely drive large employers with sicker workforces to make two types of changes. First, some employers would drop their traditional health plans and instead subsidize the purchase of individual market coverage via an HRA; this type of shifting would generally worsen the individual market risk pool. Second, some employers that do not currently offer health benefits might be induced to begin offering an HRA; this type of shifting could benefit the individual market risk pool to the extent that these employers’ sicker workers are already enrolled in the individual market. On net, as my colleagues and I discuss in detail, it appears likely that these shifts would increase individual market premiums and federal costs, although it is possible that overall insurance coverage would increase as well.

Safeguards Against Worker-Level Shifting of Sicker Workers into the Individual Market

While the Administration’s proposal as currently constructed would primarily drive the type of firm-level shifting described above, employers—including those with healthier workforces—would also have an incentive to engage in worker-level shifting: providing their sicker workers with individual market coverage via an HRA while arranging to cover their healthier workers under either a traditional health plan (in the case of large employers) or short-term, limited-duration coverage.4 Because the individual market is community-rated, while these other coverage options are not, this approach could allow employers to obtain coverage for their sicker workers at far lower cost without increasing the cost of covering their healthier workers, thereby reducing their overall cost of delivering health benefits. However, this type of health-status-based sorting would lead to substantial increases in individual market premiums, particularly given the large size of the group market relative to the individual market.

At a high level, employers could seek to achieve this sorting in two main ways. First, they could selectively offer individual market coverage to their sicker employees. Second, they could offer workers a choice between individual market coverage and other coverage, but structure that choice so as to encourage sicker workers to choose individual market coverage and healthier workers to choose the other coverage.

The Administration’s proposal includes three main provisions designed to prevent employers from selectively shifting sicker workers into the individual market. Specifically, the proposed rule: bars employers from offering the same worker a choice between this type of HRA and a traditional health plan; requires employers offering this type of HRA to offer it on the same terms to all similarly situated employees; and bars this type of HRA from being used to subsidize short-term, limited duration coverage.

These safeguards would likely be fairly effective in preventing worker-level shifting of sicker workers into the individual market, although my colleagues and I have several recommendations for how they could be strengthened. However, the proposed rule solicits comment on removing these safeguards. Removing any of these safeguards would likely cause a large influx of workers into the individual market, but the precise mechanism by which this would occur would depend on which were eliminated:

  • Allowing an HRA to be offered side-by-side with a traditional health plan: Offering workers a choice between individual market coverage subsidized via an HRA and a traditional health plan could facilitate sorting by health status. In this scenario, the employer could design its traditional health plan to be unappealing to its sicker workers, thereby encouraging those workers to decline it in favor of individual market coverage. Large employers have substantial flexibility in designing traditional health plans, so they could make these plans unappealing to sicker workers in a number of ways. They could, for example, require high cost sharing for services associated with certain high-cost chronic conditions. Another strategy would be to offer a traditional health plan that imposes high overall cost-sharing requirements, but low cost-sharing for routine care commonly used by relatively healthy workers. Employers could also selectively promote the HRA to sicker workers or otherwise encourage sicker enrollees to select the HRA option. 
  • Eliminating the requirement to offer an HRA on the same terms to all “similarly situated” employees: Employers could also selectively shift sicker workers into the individual market by only offering this type of HRA to sicker workers. Even under the proposed rule, employers would have some ability to target HRAs to sicker workers due to the large number of factors employers can use in defining what groups of workers are considered “similarly situated,” as my colleagues and I have discussed. If the list of factors was expanded, this would allow even finer-grained targeting. If these requirements were removed entirely, nearly complete sorting would likely be possible. 
  • Allowing subsidization of short-term coverage via the same HRA: If the same HRA could be used to subsidize either individual market coverage or short-term coverage, then workers’ own financial incentives would lead them to sort themselves across the two types of coverage by health status. Because short-term coverage is underwritten, sicker workers would either face high premiums or be completely unable to obtain it, so they would generally opt for individual market coverage. By contrast, healthier workers would typically be able to purchase lower-cost coverage in the short-term market and so would generally obtain their coverage there.

Methodology for Simulating the HRA Proposal Without Safeguards Against Worker-Level Shifting

I now turn to simulating the effects of a version of the Administration’s HRA proposal that lacked one or more of these safeguards against worker-level shifting. This section provides an overview of my methodology; the methodological appendix provides the full technical details. The overall effects of this version of the HRA policy would depend on two main factors: (1) how many and what types of workers end up enrolled in individual market coverage at firms that seek to use HRAs to steer their sicker workers into the individual market; and (2) how many firms elect to use HRAs in this way. I consider each in turn.

Focusing first on worker-level enrollment patterns at firms that seek to use HRAs to shift sicker workers into the individual market, I assume that a worker would end up enrolled in individual market coverage if that coverage was less expensive than (equivalent) coverage priced based on that worker’s own expected claims risk. Equivalently, I assume that people who are sicker than the individual market average (adjusting for age rating) end up enrolled in individual market coverage, while those who are healthier than the individual market average (adjusting for age rating) end up enrolled in other coverage. As discussed above, if the new type of HRA could be used to subsidize short-term coverage in addition to individual market coverage, then workers’ own choices would bring about this sorting. In policy scenarios where this type of HRA could be offered side-by-side with a traditional health plan or the requirement to offer HRAs on the same terms to all similarly situated employees were eliminated, employers would seek to bring about this sorting through their decisions about how to structure their traditional health plans or whom to offer HRAs.

In the real world, the sorting of enrollees between individual market and other coverage might be less precise than implied by this assumption because enrollment decisions might depend on factors that are not included in this simple decision rule. For example, in the policy scenario where this type of HRA could be used to purchase short-term coverage, some relatively healthy people might opt for individual market coverage even when short-term coverage would be less expensive because they wished to avoid the hassle costs associated with the underwriting process. Alternatively, some relatively sick enrollees might opt for short-term coverage (if it were available to them) despite being charged a relatively high premium because they want coverage with a broad network of providers and such coverage is unavailable in the individual market. As a crude way of capturing these and many other possible types of frictions, I report results from simulations in which the price of individual market coverage is perceived to be 25 percent higher than it actually is for the purposes of enrollment decisions, as well as simulations in which the price of individual market coverage is perceived to be 25 percent lower.

Turning to employers’ decisions, there is meaningful uncertainty regarding what share of employers would set up HRAs with the objective of shifting their sicker workers into the individual market. While the potential reduction in health benefit costs would give employers a powerful incentive to adopt this strategy, employers might reasonably fear that changing or eliminating their traditional health plans would make it harder to attract and retain workers. Indeed, depending on the precise policy scenario under consideration, setting up HRAs to achieve this type of sorting could create additional hassle costs for enrollees. Workers who are shifted into the individual market might also be displeased with the narrower networks typical of individual market products. Exactly how employers would balance these competing considerations is uncertain, so I consider scenarios in which employers accounting for 10 percent, 50 percent, or 100 percent of current enrollment in large employer coverage adopt this strategy.5 (I assume that no small employers engage in worker-level shifting, although it is likely some would in the scenario where this type of HRA could be used to purchase short-term coverage.)

To empirically implement the model sketched above, I require detailed data on the characteristics of those currently enrolled in large employer coverage. I construct a suitable sample by pooling several years of data from the Medical Expenditure Panel Survey, Household Component (MEPS-HC). In that sample, I construct a preliminary estimate of each person’s expected claims risk in the coming year based on their health care spending in the prior year, their age, and their self-reported health status. The MEPS-HC is known to understate the number of people with very high health care spending, so these estimates likely understate the number of people with very high expected health care spending. To address this problem, I adjust the preliminary estimates using estimates of variation in expected claims risk reported by Fleitas, Gowrisankaran, and Lo Sasso based on a large health care claims database.

I use an iterative method to solve for equilibrium enrollment decisions and individual market premiums. In brief, starting from current individual market premiums, I estimate the number and type of enrollees in employer coverage who would shift into the individual market. I then recompute individual market premiums based on the new enrollee pool and re-calculate which enrollees would shift into the individual market. I repeat these calculations until the resulting premium is consistent with the enrollment decisions leading to that premium. These premiums and enrollment patterns are the new equilibrium.

I note that there are at least two ways in which my approach may understate the effect of the HRA policy on individual market premiums. First, I do not model firm-level shifting into the individual market. In scenarios in which only a minority of employers engage in worker-level shifting, it is likely that some firms with sicker workforces (other than those engaged in worker-level shifting) would elect to move all of their workers into the individual market. This would likely cause individual market premiums to rise by more than I estimate here. Second, my modeling does not account for the likelihood that higher premiums would reduce enrollment among unsubsidized enrollees purchasing individual market coverage. Since those dropping individual market coverage would likely be healthier, on average, than those who remain, this too would cause individual market premiums to rise by more than I estimate here.

Simulation Results

Table 1 reports the estimated change in individual market premiums from implementing a version of the HRA proposal that lacked one or more of the safeguards against worker-level shifting included in the proposed rule. Across all of the scenarios examined, individual market premiums would increase substantially. Unsurprisingly, the amount individual market premiums would rise depends on how many employers set up HRAs designed to shift sicker workers into the individual market. If only 10 percent of employers set up such HRAs, then individual market premiums would rise by 16-17 percent, depending the set of assumptions used. By contrast, if nearly all large employers sought to use HRAs in this way, individual market premiums would almost double under all of assumptions examined.

Table 1: Change in Individual Market Premiums
Share of Large Employers Engaged in Worker-Level Shifting Effect of Factors Other Than Expected Claims Risk in Determining Worker Sorting Across Coverage Types
Perceived individual market premiums 25% lower than actual No adjustment (base assumption) Perceived individual market premium 25% higher than actual
10 percent +16.1% +17.3% +16.6%
50 percent +55.3% +59.9% +58.0%
100 percent +85.6% +93.1% +89.7%
Source: Author’s calculations using methodology described in the text.
USC Schaeffer and Brookings


Perhaps more surprising, the estimates in Table 1 show that allowing factors other than expected claims risk to influence how workers sort between the individual market and other coverage has only a modest effect of the results. There are two reasons for this. First, changes in the perceived attractiveness of individual market coverage change enrollment decisions for relatively few people. This is because average claims risk in the individual market is quite high after implementation of the HRA policy, so the level of expected claims risk where a worker (or the worker’s employer) is indifferent between individual market coverage and other coverage is well out in the tail of the distribution, where the distribution is relatively thin. Second, essentially by definition, enrollees on the margin between individual market coverage and other coverage have expected claims risk similar to the (post-policy) individual market average, so whether these enrollees are in the individual market has little effect on market average claims risk.

Table 2 shows that only a relatively small fraction of workers associated with employers engaging in worker-level shifting via an HRA would end up with individual market coverage. These shifts would nevertheless have a large effect on premiums because the employer market is very large and because those who do shift into the individual market would have very high expected claims costs. Indeed, much of the effect on individual market premiums would occur even if the number of workers shifting into the individual market was substantially smaller than shown in these simulations, provided that the very costliest enrollees at these employers did shift into the individual market.

Table 2: Share of Enrollees Shifting to Individual Market (at Shifting Firms)
Share of Large Employers Engaged in Worker-Level Shifting Effect of Factors Other Than Expected Claims Risk in Determining Worker Sorting Across Coverage Types
Perceived individual market premiums 25% lower than actual No adjustment (base assumption) Perceived individual market premium 25% higher than actual
10 percent 23.7% 15.6% 10.8%
50 percent 15.7% 9.3% 6.4%
100 percent 11.8% 6.7% 4.4%
Source: Author’s calculations using methodology described in the text.
USC Schaeffer and Brookings


Finally, Table 3 quantifies the savings employers could realize by engaging in worker-level shifting. Specifically, the table reports the change in the effective price of insurance coverage experienced by employers engaged in worker-level shifting. For these purposes, I define the change in the effective price as the percent difference between the cost of the coverage its workers would obtain under the worker-level shifting strategy (whether in the individual market or elsewhere) and the cost of delivering equivalent coverage through a traditional health plan.6

Table 3: Change in Effective Price of Coverage (for Shifting Firms)
Share of Large Employers Engaged in Worker-Level Shifting Effect of Factors Other Than Expected Claims Risk in Determining Worker Sorting Across Coverage Types
Perceived individual market premiums 25% lower than actual No adjustment (base assumption) Perceived individual market premium 25% higher than actual
10 percent -21.0% -22.6% -21.8%
50 percent -14.5% -15.7% -15.2%
100 percent -11.2% -12.2% -11.8%
Source: Author’s calculations using methodology described in the text.
USC Schaeffer and Brookings


If relatively few employers engage in worker-level shifting, then this effective price would fall by 21-23 percent, depending on the precise assumptions used. This effective price would fall by less if more employers engaged in worker-level shifting, reflecting the fact that individual market premiums would be higher. Even in these scenarios, however, engaging in worker-level shifting would reduce this effective price by 11-12 percent. The magnitude of these potential savings suggest that at least a substantial minority of firms would likely adopt this approach, despite the fact that this strategy would have some countervailing costs for employers, as discussed earlier.

Conclusion

Without the provisions of the proposed rule that aim to prevent employers from using HRAs to shift their sicker workers into the individual market, the Administration’s HRA proposal would lead to very large increases in individual market premiums, thereby increasing costs for individual market enrollees who are not eligible for subsidies and increasing the federal government’s costs of providing premium tax credits. Those costs would be difficult to justify; the main beneficiaries of the proposal would be large employers that already offer coverage, many of whom do not have particularly sick workforces, meaning that the proposal would be unlikely to produce substantial increases in insurance coverage or improvements in risk sharing. While the costs of the Administration’s HRA proposal likely outweigh its benefits even in its current form, these results imply that that if the Administration moves ahead, it should preserve—and strengthen—provisions intended to prevent selective shifting of sicker workers into the individual market.

Footnotes

  1. In this piece, I use the term “large employer” to refer to employers eligible to purchase coverage on the large group market. In most states, this category consists of employers with more than 50 employees. However, in a minority of states it consists of employers with more than 100 employees.

  2. Throughout, the term “worker” should be understood to encompass both the worker and any dependents of the worker enrolled in the employer’s plan.

  3. The Administration also proposes to create “excepted benefit” HRAs that can be used for certain other purposes not permitted under current regulations, including the purchase of short-term, limited duration insurance. This analysis does not address that portion of the Administration’s proposal.

  4. In theory, small employers, not just large employers, could implement a version of this approach in which they cover their healthier workers through a traditional health plan if they are willing to self-insure. Self-insurance has traditionally been uncommon among small employers since they are poorly positioned to bear significant claims risk. However, it is generally believed that nominally self-insured arrangements that place much less risk on employers are becoming common, which could make this a more viable strategy over the long run.

  5. Smaller shares might be appropriate in the policy scenario where this type of HRA can be used to subsidize short-term coverage since some states intend to bar or substantially limit employers’ ability to offer these policies. This would, in turn, reduce the number of employers for who worker-level shifting was an available strategy.

  6. The change in the effective price of delivering coverage similar to what the employer offered previously could be larger or smaller than the amount reported in Table 3. In particular, it is likely that the effective price of delivering coverage with a relatively broad provider network would decline by less than shown here since such coverage is often relatively expensive in the individual market (or completely unavailable). On the other hand, the effective price of delivering coverage with a narrower provider market would likely decline by more than shown here.

Get daily updates from Brookings